Download ppt

Outline







Introduction
Background
Distributed DBMS Architecture
Distributed Database Design
Semantic Data Control
Distributed Query Processing
Distributed Transaction Management
 Transaction Concepts and Models
 Distributed Concurrency Control
 Distributed Reliability




Distributed DBMS
Parallel Database Systems
Distributed Object DBMS
Database Interoperability
Concluding Remarks
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 1
Transaction
A transaction is a collection of actions that make
consistent transformations of system states while
preserving system consistency.
concurrency transparency
failure transparency
Database in a
consistent
state
Begin
Transaction
Distributed DBMS
Database may be
temporarily in an
inconsistent state
during execution
Execution of
Transaction
© 1998 M. Tamer Özsu & Patrick Valduriez
Database in a
consistent
state
End
Transaction
Page 10-12. 2
Transaction Example –
A Simple SQL Query
Transaction BUDGET_UPDATE
begin
EXEC SQL UPDATE PROJ
SET
BUDGET = BUDGET1.1
WHERE PNAME = “CAD/CAM”
end.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 3
Example Database
Consider an airline reservation example with the
relations:
FLIGHT(FNO, DATE, SRC, DEST, STSOLD, CAP)
CUST(CNAME, ADDR, BAL)
FC(FNO, DATE, CNAME,SPECIAL)
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 4
Example Transaction – SQL Version
Begin_transaction Reservation
begin
input(flight_no, date, customer_name);
EXEC SQL UPDATE FLIGHT
SET
STSOLD = STSOLD + 1
WHERE FNO = flight_no AND DATE = date;
EXEC SQL INSERT
INTO
FC(FNO, DATE, CNAME, SPECIAL);
VALUES (flight_no, date, customer_name, null);
output(“reservation completed”)
end . {Reservation}
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 5
Termination of Transactions
Begin_transaction Reservation
begin
input(flight_no, date, customer_name);
EXEC SQL SELECT
STSOLD,CAP
INTO
FROM
WHERE
if temp1 = temp2 then
temp1,temp2
FLIGHT
FNO = flight_no AND DATE = date;
output(“no free seats”);
Abort
else
EXEC SQL
EXEC SQL
UPDATE FLIGHT
SET
STSOLD = STSOLD + 1
WHERE FNO = flight_no AND DATE = date;
INSERT
INTO
FC(FNO, DATE, CNAME, SPECIAL);
VALUES (flight_no, date, customer_name, null);
Commit
output(“reservation completed”)
endif
end . {Reservation}
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 6
Example Transaction –
Reads & Writes
Begin_transaction Reservation
begin
input(flight_no, date, customer_name);
temp Read(flight_no(date).stsold);
if temp = flight(date).cap then
begin
output(“no free seats”);
Abort
end
else begin
Write(flight(date).stsold, temp + 1);
Write(flight(date).cname, customer_name);
Write(flight(date).special, null);
Commit;
output(“reservation completed”)
end
end. {Reservation}
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 7
Characterization

Read set (RS)
 The set of data items that are read by a transaction

Write set (WS)
 The set of data items whose values are changed by
this transaction

Base set (BS)
 RS  WS
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 8
Formalization
Let
 Oij(x) be some operation Oj of transaction Ti operating on
entity x, where Oj  {read,write} and Oj is atomic
 OSi = j Oij
 Ni  {abort,commit}
Transaction Ti is a partial order Ti = {i, <i} where
 i = OSi {Ni }
 For any two operations Oij , Oik OSi , if Oij = R(x)
and Oik = W(x) for any data item x, then either
Oij <i Oik or Oik <i Oij
 Oij OSi, Oij <i Ni
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 9
Example
Consider a transaction T:
Read(x)
Read(y)
x x + y
Write(x)
Commit
Then
 = {R(x), R(y), W(x), C}
< = {(R(x), W(x)), (R(y), W(x)), (W(x), C), (R(x), C), (R(y), C)}
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 10
DAG Representation
Assume
< = {(R(x),W(x)), (R(y),W(x)), (R(x), C), (R(y), C), (W(x), C)}
R(x)
W(x)
C
R(y)
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 11
Properties of Transactions
ATOMICITY
 all or nothing
CONSISTENCY
 no violation of integrity constraints
ISOLATION
 concurrent changes invisible È serializable
DURABILITY
 committed updates persist
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 12
Atomicity




Either all or none of the transaction's operations are
performed.
Atomicity requires that if a transaction is
interrupted by a failure, its partial results must be
undone.
The activity of preserving the transaction's atomicity
in presence of transaction aborts due to input errors,
system overloads, or deadlocks is called transaction
recovery.
The activity of ensuring atomicity in the presence of
system crashes is called crash recovery.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 13
Consistency

Internal consistency
 A transaction which executes alone against a
consistent database leaves it in a consistent state.
 Transactions do not violate database integrity
constraints.

Distributed DBMS
Transactions are correct programs
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 14
Consistency Degrees

Degree 0
 Transaction T does not overwrite dirty data of other
transactions
 Dirty data refers to data values that have been
updated by a transaction prior to its commitment

Degree 1
 T does not overwrite dirty data of other transactions
 T does not commit any writes before EOT
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 15
Consistency Degrees (cont’d)

Degree 2
 T does not overwrite dirty data of other transactions
 T does not commit any writes before EOT
 T does not read dirty data from other transactions

Degree 3
 T does not overwrite dirty data of other transactions
 T does not commit any writes before EOT
 T does not read dirty data from other transactions
 Other transactions do not dirty any data read by T
before T completes.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 16
Isolation

Serializability
 If several transactions are executed concurrently,
the results must be the same as if they were
executed serially in some order.

Incomplete results
 An incomplete transaction cannot reveal its results
to other transactions before its commitment.
 Necessary to avoid cascading aborts.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 17
Isolation Example


Consider the following two transactions:
T1: Read(x)
T2: Read(x)
x x1
x x1
Write(x)
Write(x)
Commit
Commit
Possible execution sequences:
T1:
T1:
T1:
T1:
T2:
T2:
T2:
T2:
Distributed DBMS
Read(x)
x x1
Write(x)
Commit
Read(x)
x x1
Write(x)
Commit
T1:
T1:
T2:
T1:
T2:
T2:
T1:
T2:
Read(x)
x x1
Read(x)
Write(x)
x x1
Write(x)
Commit
Commit
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 18
SQL-92 Isolation Levels
Phenomena:
 Dirty read
 T1 modifies x which is then read by T2 before T1
terminates; T1 aborts  T2 has read value which
never exists in the database.

Non-repeatable (fuzzy) read
 T1 reads x; T2 then modifies or deletes x and
commits. T1 tries to read x again but reads a
different value or can’t find it.

Phantom
 T1 searches the database according to a predicate
while T2 inserts new tuples that satisfy the
predicate.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 19
SQL-92 Isolation Levels (cont’d)

Read Uncommitted
 For transactions operating at this level, all three
phenomena are possible.

Read Committed
 Fuzzy reads and phantoms are possible, but dirty
reads are not.

Repeatable Read
 Only phantoms possible.

Anomaly Serializable
 None of the phenomena are possible.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 20
Durability

Once a transaction commits, the system
must guarantee that the results of its
operations will never be lost, in spite of
subsequent failures.

Database recovery
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 21
Characterization of Transactions
Based on
 Application areas
 non-distributed vs. distributed
compensating transactions
 heterogeneous transactions

 Timing

on-line (short-life) vs batch (long-life)
 Organization of read and write actions
two-step
 restricted
 action model

 Structure
flat (or simple) transactions
 nested transactions
 workflows

Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 22
Transaction Structure

Flat transaction
 Consists of a sequence of primitive operations embraced
between a begin and end markers.
Begin_transaction Reservation
…
end.

Nested transaction
 The operations of a transaction may themselves be
transactions.
Begin_transaction Reservation
…
Begin_transaction Airline
– …
end. {Airline}
Begin_transaction Hotel
…
end. {Hotel}
end. {Reservation}
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 23
Nested Transactions

Have the same properties as their parents  may
themselves have other nested transactions.

Introduces concurrency control and recovery
concepts to within the transaction.

Types
 Closed nesting

Subtransactions begin after their parents and finish before
them.

Commitment of a subtransaction is conditional upon the
commitment of the parent (commitment through the root).
 Open nesting
Distributed DBMS

Subtransactions can execute and commit independently.

Compensation may be necessary.
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 24
Workflows


“A collection of tasks organized to accomplish some
business process.” [D. Georgakopoulos]
Types
 Human-oriented workflows
Involve humans in performing the tasks.
 System support for collaboration and coordination; but no
system-wide consistency definition

 System-oriented workflows
Computation-intensive & specialized tasks that can be
executed by a computer
 System support for concurrency control and recovery,
automatic task execution, notification, etc.

 Transactional workflows

Distributed DBMS
In between the previous two; may involve humans, require
access to heterogeneous, autonomous and/or distributed
systems, and support selective use of ACID properties
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 25
Workflow Example
T3
T1
T2
Customer
Database
Customer
Database
Distributed DBMS
T5
T4
T1: Customer request
obtained
T2: Airline reservation
performed
T3: Hotel reservation
performed
T4: Auto reservation
performed
T5: Bill generated
Customer
Database
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 26
Transactions Provide…

Atomic and reliable execution in the presence
of failures

Correct execution in the presence of multiple
user accesses

Correct management of replicas (if they
support it)
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 27
Transaction Processing Issues

Transaction structure (usually called
transaction model)
 Flat (simple), nested

Internal database consistency
 Semantic data control (integrity enforcement)
algorithms

Reliability protocols
 Atomicity & Durability
 Local recovery protocols
 Global commit protocols
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 28
Transaction Processing Issues

Concurrency control algorithms
 How to synchronize concurrent transaction
executions (correctness criterion)
 Intra-transaction consistency, Isolation

Replica control protocols
 How to control the mutual consistency of replicated
data
 One copy equivalence and ROWA
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 29
Architecture Revisited
Begin_transaction,
Read, Write,
Commit, Abort
Results
Distributed
Execution Monitor
With other
TMs
Transaction Manager
(TM)
Scheduling/
Descheduling
Requests
Scheduler
(SC)
With other
SCs
To data
processor
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 30
Centralized Transaction Execution
User
Application
User
Application
…
Begin_Transaction,
Read, Write, Abort, EOT
Results &
User Notifications
Transaction
Manager
(TM)
Read, Write,
Abort, EOT
Results
Scheduler
(SC)
Scheduled
Operations
Results
Recovery
Manager
(RM)
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 31
Distributed Transaction Execution
User application
Results &
User notifications
Begin_transaction,
Read, Write, EOT,
Abort
TM
Distributed
Transaction Execution
Model
TM
Replica Control
Protocol
Read, Write,
EOT, Abort
SC
RM
Distributed DBMS
SC
Distributed
Concurrency Control
Protocol
RM
Local
Recovery
Protocol
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 32
Concurrency Control

The problem of synchronizing concurrent
transactions such that the consistency of the
database is maintained while, at the same
time, maximum degree of concurrency is
achieved.

Anomalies:
 Lost updates

The effects of some transactions are not reflected on
the database.
 Inconsistent retrievals

Distributed DBMS
A transaction, if it reads the same data item more than
once, should always read the same value.
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 33
Execution Schedule (or History)


An order in which the operations of a set of
transactions are executed.
A schedule (history) can be defined as a partial
order over the operations of a set of transactions.
T1: Read(x)
Write(x)
Commit
T2: Write(x)
Write(y)
Read(z)
Commit
T3: Read(x)
Read(y)
Read(z)
Commit
H1={W2(x),R1(x), R3(x),W1(x),C1,W2(y),R3(y),R2(z),C2,R3(z),C3}
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 34
Formalization of Schedule
A complete schedule SC(T) over a set of
transactions T={T1, …, Tn} is a partial order
SC(T)={T, < T} where
 T = i i , for i = 1, 2, …, n
 < T i < i , for i = 1, 2, …, n
 For any two conflicting operations Oij, Okl  T,
either Oij < T Okl or Okl < T Oij
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 35
Complete Schedule – Example
Given three transactions
T1: Read(x)
Write(x)
Commit
T2: Write(x)
Write(y)
Read(z)
Commit
T3: Read(x)
Read(y)
Read(z)
Commit
A possible complete schedule is given as the DAG
Distributed DBMS
R1(x)
W2(x)
R3(x)
W1(x)
W2(y)
R3(y)
C1
R2(z)
R3(z)
C2
C3
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 36
Schedule Definition
A schedule is a prefix of a complete schedule
such that only some of the operations and only
some of the ordering relationships are included.
T1: Read(x)
Write(x)
Commit
T2: Write(x)
Write(y)
Read(z)
Commit
R1(x)
W2(x)
R3(x)
W1(x)
W2(y)
R3(y)
C1
R2(z)
R3(z)
C2
C3
Distributed DBMS
T3: Read(x)
Read(y)
Read(z)
Commit
R1(x)

© 1998 M. Tamer Özsu & Patrick Valduriez
W2(x)
R3(x)
W2(y)
R3(y)
R2(z)
R3(z)
Page 10-12. 37
Serial History



All the actions of a transaction occur
consecutively.
No interleaving of transaction operations.
If each transaction is consistent (obeys
integrity rules), then the database is
guaranteed to be consistent at the end of
executing a serial history.
T1: Read(x)
Write(x)
Commit
T2: Write(x)
Write(y)
Read(z)
Commit
T3: Read(x)
Read(y)
Read(z)
Commit
Hs={W2(x),W2(y),R2(z),C2,R1(x),W1(x),C1,R3(x),R3(y),R3(z),C3}
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 38
Serializable History


Transactions execute concurrently, but the net
effect of the resulting history upon the database
is equivalent to some serial history.
Equivalent with respect to what?
 Conflict equivalence: the relative order of
execution of the conflicting operations belonging to
unaborted transactions in two histories are the
same.
 Conflicting operations: two incompatible
operations (e.g., Read and Write) conflict if they both
access the same data item.
Incompatible operations of each transaction is assumed
to conflict; do not change their execution orders.
 If two operations from two different transactions
conflict, the corresponding transactions are also said to
conflict.

Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 39
Serializable History
T1: Read(x)
Write(x)
Commit
T2: Write(x)
Write(y)
Read(z)
Commit
T3: Read(x)
Read(y)
Read(z)
Commit
The following are not conflict equivalent
Hs={W2(x),W2(y),R2(z),C2,R1(x),W1(x),C1,R3(x),R3(y),R3(z),C3}
H1={W2(x),R1(x), R3(x),W1(x),C1,W2(y),R3(y),R2(z),C2,R3(z),C3}
The following are conflict equivalent; therefore
H2 is serializable.
Hs={W2(x),W2(y),R2(z),C2,R1(x),W1(x),C1,R3(x),R3(y),R3(z),C3}
H2={W2(x),R1(x),W1(x),C1,R3(x),W2(y),R3(y),R2(z),C2,R3(z),C3}
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 40
Serializability in Distributed DBMS

Somewhat more involved. Two histories have to be
considered:
 local histories
 global history

For global transactions (i.e., global history) to be
serializable, two conditions are necessary:
 Each local history should be serializable.
 Two conflicting operations should be in the same relative
order in all of the local histories where they appear together.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 41
Global Non-serializability
T1: Read(x)
x x5
Write(x)
Commit
T2: Read(x)
x x15
Write(x)
Commit
The following two local histories are individually
serializable (in fact serial), but the two transactions
are not globally serializable.
LH1={R1(x),W1(x),C1,R2(x),W2(x),C2}
LH2={R2(x),W2(x),C2,R1(x),W1(x),C1}
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 42
Concurrency Control
Algorithms

Pessimistic
 Two-Phase Locking-based (2PL)
Centralized (primary site) 2PL
 Primary copy 2PL
 Distributed 2PL

 Timestamp Ordering (TO)
Basic TO
 Multiversion TO
 Conservative TO

 Hybrid

Optimistic
 Locking-based
 Timestamp ordering-based
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 43
Locking-Based Algorithms




Transactions indicate their intentions by
requesting locks from the scheduler (called lock
manager).
Locks are either read lock (rl) [also called shared
lock] or write lock (wl) [also called exclusive lock]
Read locks and write locks conflict (because Read
and Write operations are incompatible
rl
wl
rl
yes no
wl
no
no
Locking works nicely to allow concurrent
processing of transactions.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 44
Two-Phase Locking (2PL)
 A Transaction locks an object before using it.
 When an object is locked by another transaction,
the requesting transaction must wait.
 When a transaction releases a lock, it may not
request another lock.
Lock point
Obtain lock
No. of locks
Release lock
Phase 1
Phase 2
BEGIN
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
END
Page 10-12. 45
Strict 2PL
Hold locks until the end.
Obtain lock
Release lock
BEGIN
END
period of
data item
use
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Transaction
duration
Page 10-12. 46
Centralized 2PL


There is only one 2PL scheduler in the distributed system.
Lock requests are issued to the central scheduler.
Data Processors at
participating sites
Distributed DBMS
Coordinating TM
Central Site LM
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 47
Distributed 2PL

2PL schedulers are placed at each site. Each
scheduler handles lock requests for data at that site.

A transaction may read any of the replicated copies
of item x, by obtaining a read lock on one of the
copies of x. Writing into x requires obtaining write
locks for all copies of x.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 48
Distributed 2PL Execution
Coordinating TM
Distributed DBMS
Participating LMs
© 1998 M. Tamer Özsu & Patrick Valduriez
Participating DPs
Page 10-12. 49
Timestamp Ordering
Transaction (Ti) is assigned a globally unique
timestamp ts(Ti).
Transaction manager attaches the timestamp to all
operations issued by the transaction.
Each data item is assigned a write timestamp (wts) and
a read timestamp (rts):
rts(x) = largest timestamp of any read on x
wts(x) = largest timestamp of any read on x
Conflicting operations are resolved by timestamp order.
Basic T/O:
for Ri(x)
if ts(Ti) < wts(x)
then reject Ri(x)
else accept Ri(x)
rts(x) ts(Ti)
Distributed DBMS
for Wi(x)
if ts(Ti) < rts(x) and ts(Ti) < wts(x)
then reject Wi(x)
else accept Wi(x)
wts(x) ts(Ti)
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 50
Conservative Timestamp
Ordering

Basic timestamp ordering tries to
execute an operation as soon as it
receives it
 progressive
 too many restarts since there is no delaying


Conservative timestamping delays each
operation until there is an assurance
that it will not be restarted
Assurance?
 No other operation with a smaller
timestamp can arrive at the scheduler
 Note that the delay may result in the
formation of deadlocks
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 51
Multiversion Timestamp Ordering

Do not modify the values in the database,
create new values.

A Ri(x) is translated into a read on one version
of x.
 Find a version of x (say xv) such that ts(xv) is the
largest timestamp less than ts(Ti).

A Wi(x) is translated into Wi(xw) and accepted if
the scheduler has not yet processed any Rj(xr)
such that
ts(Ti) < ts(xr) < ts(Tj)
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 52
Optimistic Concurrency Control
Algorithms
Pessimistic execution
Validate
Read
Compute
Write
Validate
Write
Optimistic execution
Read
Distributed DBMS
Compute
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 53
Optimistic Concurrency Control
Algorithms

Transaction execution model: divide into
subtransactions each of which execute at a site
 Tij: transaction Ti that executes at site j

Transactions run independently at each site
until they reach the end of their read phases

All subtransactions are assigned a timestamp
at the end of their read phase

Validation test performed during validation
phase. If one fails, all rejected.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 54
Optimistic CC Validation Test
 If all transactions Tk where ts(Tk) < ts(Tij)
have completed their write phase before Tij
has started its read phase, then validation
succeeds
 Transaction executions in serial order
Tk
R
V
W
Tij
Distributed DBMS
R
© 1998 M. Tamer Özsu & Patrick Valduriez
V
W
Page 10-12. 55
Optimistic CC Validation Test
 If there is any transaction Tk such that ts(Tk)<ts(Tij)
and which completes its write phase while Tij is in
its read phase, then validation succeeds if
WS(Tk)  RS(Tij) = Ø
 Read and write phases overlap, but Tij does not read data
items written by Tk
Tk
R
V
W
Tij
Distributed DBMS
R
V
© 1998 M. Tamer Özsu & Patrick Valduriez
W
Page 10-12. 56
Optimistic CC Validation Test
 If there is any transaction Tk such that ts(Tk)< ts(Tij)
and which completes its read phase before Tij
completes its read phase, then validation succeeds if
WS(Tk) RS(Tij) = Ø and WS(Tk) WS(Tij) = Ø
 They overlap, but don't access any common data items.
R
Tk
Tij
Distributed DBMS
V
R
W
V
W
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 57
Deadlock

A transaction is deadlocked if it is blocked and will
remain blocked until there is intervention.

Locking-based CC algorithms may cause deadlocks.

TO-based algorithms that involve waiting may cause
deadlocks.

Wait-for graph
 If transaction Ti waits for another transaction Tj to release
a lock on an entity, then Ti  Tj in WFG.
Ti
Distributed DBMS
Tj
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 58
Local versus Global WFG
Assume T1 and T2 run at site 1, T3 and T4 run at site 2.
Also assume T3 waits for a lock held by T4 which waits
for a lock held by T1 which waits for a lock held by T2
which, in turn, waits for a lock held by T3.
Local WFG Site 1
Site 2
T1
T4
T2
T3
Global WFG
Distributed DBMS
T1
T4
T2
T3
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 59
Deadlock Management

Ignore
 Let the application programmer deal with it, or
restart the system

Prevention
 Guaranteeing that deadlocks can never occur in
the first place. Check transaction when it is
initiated. Requires no run time support.

Avoidance
 Detecting potential deadlocks in advance and
taking action to insure that deadlock will not
occur. Requires run time support.

Detection and Recovery
 Allowing deadlocks to form and then finding and
breaking them. As in the avoidance scheme, this
requires run time support.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 60
Deadlock Prevention

All resources which may be needed by a transaction
must be predeclared.
 The system must guarantee that none of the resources will
be needed by an ongoing transaction.
 Resources must only be reserved, but not necessarily
allocated a priori
 Unsuitability of the scheme in database environment
 Suitable for systems that have no provisions for undoing
processes.

Evaluation:
– Reduced concurrency due to preallocation
– Evaluating whether an allocation is safe leads to added
overhead.
– Difficult to determine (partial order)
+ No transaction rollback or restart is involved.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 61
Deadlock Avoidance

Transactions are not required to request
resources a priori.

Transactions are allowed to proceed unless a
requested resource is unavailable.

In case of conflict, transactions may be
allowed to wait for a fixed time interval.

Order either the data items or the sites and
always request locks in that order.

More attractive than prevention in a
database environment.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 62
Deadlock Avoidance –
Wait-Die & Wound-Wait Algorithms
WAIT-DIE Rule: If Ti requests a lock on a data item
which is already locked by Tj, then Ti is permitted to
wait iff ts(Ti)<ts(Tj). If ts(Ti)>ts(Tj), then Ti is aborted
and restarted with the same timestamp.
 if ts(Ti)<ts(Tj) then Ti waits else Ti dies
 non-preemptive: Ti never preempts Tj
 prefers younger transactions
WOUND-WAIT Rule: If Ti requests a lock on a data
item which is already locked by Tj , then Ti is
permitted to wait iff ts(Ti)>ts(Tj). If ts(Ti)<ts(Tj), then
Tj is aborted and the lock is granted to Ti.
 if ts(Ti)<ts(Tj) then Tj is wounded else Ti waits
 preemptive: Ti preempts Tj if it is younger
 prefers older transactions
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 63
Deadlock Detection

Transactions are allowed to wait freely.

Wait-for graphs and cycles.

Topologies for deadlock detection
algorithms
 Centralized
 Distributed
 Hierarchical
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 64
Centralized Deadlock Detection

One site is designated as the deadlock detector for
the system. Each scheduler periodically sends its
local WFG to the central site which merges them to
a global WFG to determine cycles.

How often to transmit?
 Too often  higher communication cost but lower delays
due to undetected deadlocks
 Too late  higher delays due to deadlocks, but lower
communication cost

Would be a reasonable choice if the concurrency
control algorithm is also centralized.

Proposed for Distributed INGRES
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 65
Hierarchical Deadlock Detection
Build a hierarchy of detectors
DDox
DD11
Site 1
DD21
Distributed DBMS
DD14
Site 2 Site 3
DD22
DD23
© 1998 M. Tamer Özsu & Patrick Valduriez
Site 4
DD24
Page 10-12. 66
Distributed Deadlock Detection


Sites cooperate in detection of deadlocks.
One example:
 The local WFGs are formed at each site and passed on to
other sites. Each local WFG is modified as follows:
 Since each site receives the potential deadlock cycles from
other sites, these edges are added to the local WFGs
 The edges in the local WFG which show that local
transactions are waiting for transactions at other sites are
joined with edges in the local WFGs which show that remote
transactions are waiting for local ones.
 Each local deadlock detector:


Distributed DBMS
looks for a cycle that does not involve the external edge. If it
exists, there is a local deadlock which can be handled locally.
looks for a cycle involving the external edge. If it exists, it
indicates a potential global deadlock. Pass on the information
to the next site.
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 67
Reliability
Problem:
How to maintain
atomicity
durability
properties of transactions
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 68
Fundamental Definitions

Reliability
 A measure of success with which a system conforms
to some authoritative specification of its behavior.
 Probability that the system has not experienced any
failures within a given time period.
 Typically used to describe systems that cannot be
repaired or where the continuous operation of the
system is critical.

Availability
 The fraction of the time that a system meets its
specification.
 The probability that the system is operational at a
given time t.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 69
Basic System Concepts
ENVIRONMENT
SYSTEM
Component 1
Component 2
Stimuli
Responses
Component 3
External state
Internal state
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 70
Fundamental Definitions

Failure
 The deviation of a system from the behavior that is
described in its specification.

Erroneous state
 The internal state of a system such that there exist
circumstances in which further processing, by the
normal algorithms of the system, will lead to a
failure which is not attributed to a subsequent fault.

Error
 The part of the state which is incorrect.

Fault
 An error in the internal states of the components of
a system or in the design of a system.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 71
Faults to Failures
causes
Fault
Distributed DBMS
results in
Error
© 1998 M. Tamer Özsu & Patrick Valduriez
Failure
Page 10-12. 72
Types of Faults

Hard faults
 Permanent
 Resulting failures are called hard failures

Soft faults
 Transient or intermittent
 Account for more than 90% of all failures
 Resulting failures are called soft failures
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 73
Fault Classification
Permanent
fault
Permanent
error
Incorrect
design
Unstable or
marginal
components
Unstable
environment
Intermittent
error
System
Failure
Transient
error
Operator
mistake
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 74
Failures
MTBF
MTTD
MTTR
Time
Fault
Error
occurs caused
Detection
of error
Repair
Fault Error
occurs caused
Multiple errors can occur
during this period
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 75
Fault Tolerance Measures
Reliability
R(t) = Pr{0 failures in time [0,t] | no failures at t=0}
If occurrence of failures is Poisson
R(t) = Pr{0 failures in time [0,t]}
Then
Pr(k failures in time [0,t] =
e-m(t)[m(t)]k
k!
where m(t) is known as the hazard function which
gives the time-dependent failure rate of the
component and is defined as

t
m(t)  z(x)dx
0
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 76
Fault-Tolerance Measures
Reliability
The mean number of failures in time [0, t] can be
computed as
∞
e-m(t )[m(t )]k
E [k] = k
= m(t )
k!
k =0
and the variance can be be computed as
Var[k] = E[k2] - (E[k])2 = m(t)
Thus, reliability of a single component is
R(t) = e-m(t)
and of a system consisting of n non-redundant
components as
n
Rsys(t) =  Ri(t)
i =1
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 77
Fault-Tolerance Measures
Availability
A(t) = Pr{system is operational at time t}
Assume

Poisson failures with rate

Repair time is exponentially distributed with mean 1/µ
Then, steady-state availability
A = lim A(t) 
t 
Distributed DBMS


© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 78
Fault-Tolerance Measures
MTBF
Mean time between failures
MTBF =  R(t)dt
∞
MTTR
Mean time to repair
Availability
MTBF
MTBF + MTTR
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 79
Sources of Failure –
SLAC Data (1985)
Software
13%
Operations
57%
Hardware
13%
Environment
17%
S. Mourad and D. Andrews, “The Reliability of the IBM/XA
Operating System”, Proc. 15th Annual Int. Symp. on FTCS, 1985.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 80
Sources of Failure –
Japanese Data (1986)
Operations
10% Environment
11%
Application
SW
25%
Vendor
42%
Comm.
Lines
12%
“Survey on Computer Security”, Japan Info. Dev. Corp.,1986.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 81
Sources of Failure –
5ESS Switch (1987)
Operations
18%
Software
44%
Unknown
6%
Hardware
32%
D.A. Yaeger. 5ESS Switch Performance Metrics. Proc. Int.
Conf. on Communications, Volume 1, pp. 46-52, June 1987.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 82
Sources of Failures –
Tandem Data (1985)
Maintenance
25%
Operations
17%
Environment
14%
Hardware
18%
Software
26%
Jim Gray, Why Do Computers Stop and What can be
Done About It?, Tandem Technical Report 85.7, 1985.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 83
Types of Failures

Transaction failures
 Transaction aborts (unilaterally or due to deadlock)
 Avg. 3% of transactions abort abnormally

System (site) failures
 Failure of processor, main memory, power supply, …
 Main memory contents are lost, but secondary storage
contents are safe
 Partial vs. total failure

Media failures
 Failure of secondary storage devices such that the
stored data is lost
 Head crash/controller failure (?)

Communication failures
 Lost/undeliverable messages
 Network partitioning
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 84
Local Recovery Management –
Architecture

Volatile storage
 Consists of the main memory of the computer system
(RAM).

Stable storage
 Resilient to failures and loses its contents only in the
presence of media failures (e.g., head crashes on disks).
 Implemented via a combination of hardware (non-volatile
storage) and software (stable-write, stable-read, clean-up)
components.
Secondary
storage
Local Recovery
Manager
Main memory
Fetch,
Flush
Read
Stable
database
Write
Distributed DBMS
Write
Database Buffer
Manager
Read
© 1998 M. Tamer Özsu & Patrick Valduriez
Database
buffers
(Volatile
database)
Page 10-12. 85
Update Strategies

In-place update
 Each update causes a change in one or more data
values on pages in the database buffers

Out-of-place update
 Each update causes the new value(s) of data item(s)
to be stored separate from the old value(s)
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 86
In-Place Update Recovery
Information
Database Log
Every action of a transaction must not only perform
the action, but must also write a log record to an
append-only file.
Old
stable database
state
Update
Operation
New
stable database
state
Database
Log
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 87
Logging
The log contains information used by the
recovery process to restore the consistency of a
system. This information may include
 transaction identifier
 type of operation (action)
 items accessed by the transaction to perform the
action
 old value (state) of item (before image)
 new value (state) of item (after image)
…
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 88
Why Logging?
Upon recovery:
 all of T1's effects should be reflected in the database
(REDO if necessary due to a failure)
 none of T2's effects should be reflected in the
database (UNDO if necessary)
system
crash
Begin
Begin
0
Distributed DBMS
T1
End
T2
t
© 1998 M. Tamer Özsu & Patrick Valduriez
time
Page 10-12. 89
REDO Protocol
Old
stable database
state
REDO
New
stable database
state
Database
Log



REDO'ing an action means performing it again.
The REDO operation uses the log information
and performs the action that might have been
done before, or not done due to failures.
The REDO operation generates the new image.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 90
UNDO Protocol
New
stable database
state
UNDO
Old
stable database
state
Database
Log


UNDO'ing an action means to restore the
object to its before image.
The UNDO operation uses the log information
and restores the old value of the object.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 91
When to Write Log Records
Into Stable Store
Assume a transaction T updates a page P
 Fortunate case
 System writes P in stable database
 System updates stable log for this update
 SYSTEM FAILURE OCCURS!... (before T commits)

We can recover (undo) by restoring P to its old state
by using the log
Unfortunate case
 System writes P in stable database
 SYSTEM FAILURE OCCURS!... (before stable log is
updated)

We cannot recover from this failure because there is
no log record to restore the old value.
Solution: Write-Ahead Log (WAL) protocol
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 92
Write–Ahead Log Protocol

Notice:
 If a system crashes before a transaction is committed,
then all the operations must be undone. Only need the
before images (undo portion of the log).
 Once a transaction is committed, some of its actions
might have to be redone. Need the after images (redo
portion of the log).

WAL protocol :
 Before a stable database is updated, the undo portion of
the log should be written to the stable log
 When a transaction commits, the redo portion of the log
must be written to stable log prior to the updating of
the stable database.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 93
Logging Interface
Secondary
storage
Main memory
Stable
log
Local Recovery
Manager
Stable
database
Fetch,
Flush
Database Buffer
Manager
Distributed DBMS
Read
Write
Log
buffers
Read
Write
© 1998 M. Tamer Özsu & Patrick Valduriez
Database
buffers
(Volatile
database)
Page 10-12. 94
Out-of-Place Update
Recovery Information

Shadowing
 When an update occurs, don't change the old page, but
create a shadow page with the new values and write it
into the stable database.
 Update the access paths so that subsequent accesses
are to the new shadow page.
 The old page retained for recovery.

Differential files
 For each file F maintain
 a read only part FR


a differential file consisting of insertions part DF+ and
deletions part DFThus, F = (FR  DF+) – DF-
 Updates treated as delete old value, insert new value
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 95
Execution of Commands
Commands to consider:
begin_transaction
read
write
commit
abort
recover
Distributed DBMS
Independent of execution
strategy for LRM
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 96
Execution Strategies

Dependent upon
 Can the buffer manager decide to write some of
the buffer pages being accessed by a transaction
into stable storage or does it wait for LRM to
instruct it?

fix/no-fix decision
 Does the LRM force the buffer manager to write
certain buffer pages into stable database at the
end of a transaction's execution?


flush/no-flush decision
Possible execution strategies:
 no-fix/no-flush
 no-fix/flush
 fix/no-flush
 fix/flush
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 97
No-Fix/No-Flush

Abort
 Buffer manager may have written some of the updated
pages into stable database
 LRM performs transaction undo (or partial undo)

Commit
 LRM writes an “end_of_transaction” record into the log.

Recover
 For those transactions that have both a
“begin_transaction” and an “end_of_transaction” record
in the log, a partial redo is initiated by LRM
 For those transactions that only have a
“begin_transaction” in the log, a global undo is executed
by LRM
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 98
No-Fix/Flush

Abort
 Buffer manager may have written some of the
updated pages into stable database
 LRM performs transaction undo (or partial undo)

Commit
 LRM issues a flush command to the buffer
manager for all updated pages
 LRM writes an “end_of_transaction” record into the
log.

Recover
 No need to perform redo
 Perform global undo
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 99
Fix/No-Flush

Abort
 None of the updated pages have been written
into stable database
 Release the fixed pages

Commit
 LRM writes an “end_of_transaction” record into
the log.
 LRM sends an unfix command to the buffer
manager for all pages that were previously
fixed

Recover
 Perform partial redo
 No need to perform global undo
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 100
Fix/Flush

Abort
 None of the updated pages have been written into stable
database
 Release the fixed pages

Commit (the following have to be done atomically)
 LRM issues a flush command to the buffer manager for
all updated pages
 LRM sends an unfix command to the buffer manager
for all pages that were previously fixed
 LRM writes an “end_of_transaction” record into the log.

Recover
 No need to do anything
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 101
Checkpoints

Simplifies the task of determining actions of
transactions that need to be undone or
redone when a failure occurs.

A checkpoint record contains a list of active
transactions.

Steps:
 Write a begin_checkpoint record into the log
 Collect the checkpoint dat into the stable storage
 Write an end_checkpoint record into the log
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 102
Media Failures –
Full Architecture
Secondary
storage
Main memory
Stable
log
Local Recovery
Manager
Stable
database
Fetch,
Flush
Database Buffer
Manager
Read
Write
Write
Archive
database
Distributed DBMS
Log
buffers
Read
Write
Database
buffers
(Volatile
database)
Write
Archive
log
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 103
Distributed Reliability Protocols

Commit protocols
 How to execute commit command for distributed
transactions.
 Issue: how to ensure atomicity and durability?

Termination protocols
 If a failure occurs, how can the remaining operational
sites deal with it.
 Non-blocking : the occurrence of failures should not force
the sites to wait until the failure is repaired to terminate
the transaction.

Recovery protocols
 When a failure occurs, how do the sites where the failure
occurred deal with it.
 Independent : a failed site can determine the outcome of a
transaction without having to obtain remote information.

Independent recovery  non-blocking termination
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 104
Two-Phase Commit (2PC)
Phase 1 : The coordinator gets the participants
ready to write the results into the database
Phase 2 : Everybody writes the results into the
database
 Coordinator :The process at the site where the
transaction originates and which controls the
execution
 Participant :The process at the other sites that
participate in executing the transaction
Global Commit Rule:
 The coordinator aborts a transaction if and only if at
least one participant votes to abort it.
 The coordinator commits a transaction if and only if
all of the participants vote to commit it.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 105
Centralized 2PC
P
P
P
P
C
C
P
P
P
P
ready?
yes/no
Phase 1
Distributed DBMS
C
commit/abort? commited/aborted
Phase 2
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 106
2PC Protocol Actions
Coordinator
Participant
INITIAL
INITIAL
write
begin_commit
in log
write abort
in log
No
Ready to
Commit?
Yes
VOTE-COMMIT
WAIT
Yes
Any No?
write ready
in log
GLOBAL-ABORT
write abort
in log
READY
No
write commit
in log
Abort
ACK
COMMIT
ABORT
ACK
write
end_of_transaction
in log
Distributed DBMS
write abort
in log
Type of
msg
Commit
write commit
in log
ABORT
© 1998 M. Tamer Özsu & Patrick Valduriez
COMMIT
Page 10-12. 107
Linear 2PC
Phase 1
Prepare
1
VC/VA
2
GC/GA
VC/VA
3
GC/GA
VC/VA
4
GC/GA
VC/VA
5
GC/GA
N
GC/GA
Phase 2
VC: Vote-Commit, VA: Vote-Abort, GC: Global-commit, GA: Global-abort
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 108
Distributed 2PC
Coordinator
Participants
Participants
global-commit/
global-abort
vote-abort/ decision made
vote-commit independently
prepare
Phase 1
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 109
State Transitions in 2PC
INITIAL
INITIAL
Commit command
Prepare
Prepare
Vote-commit
Prepare
Vote-abort
WAIT
Vote-abort
Global-abort
READY
Vote-commit (all)
Global-commit
ABORT
COMMIT
Global-abort
Ack
ABORT
Coordinator
Distributed DBMS
Global-commit
Ack
© 1998 M. Tamer Özsu & Patrick Valduriez
COMMIT
Participants
Page 10-12. 110
Site Failures - 2PC Termination
COORDINATOR

Timeout in INITIAL
 Who cares

INITIAL
Timeout in WAIT
 Cannot unilaterally commit
Commit command
Prepare
 Can unilaterally abort

Timeout in ABORT or COMMIT
 Stay blocked and wait for the acks
WAIT
Vote-abort
Global-abort
ABORT
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Vote-commit
Global-commit
COMMIT
Page 10-12. 111
Site Failures - 2PC Termination
PARTICIPANTS

INITIAL
Timeout in INITIAL
 Coordinator must have
failed in INITIAL state
 Unilaterally abort

Prepare
Vote-commit
Prepare
Vote-abort
Timeout in READY
 Stay blocked
READY
Global-abort
Ack
ABORT
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Global-commit
Ack
COMMIT
Page 10-12. 112
Site Failures - 2PC Recovery
COORDINATOR

Failure in INITIAL
INITIAL
 Start the commit process upon recovery

Failure in WAIT
 Restart the commit process upon
Commit command
Prepare
recovery

WAIT
Failure in ABORT or COMMIT
 Nothing special if all the acks have
been received
 Otherwise the termination protocol is
involved
Distributed DBMS
Vote-abort
Global-abort
© 1998 M. Tamer Özsu & Patrick Valduriez
ABORT
Vote-commit
Global-commit
COMMIT
Page 10-12. 113
Site Failures - 2PC Recovery
PARTICIPANTS

Failure in INITIAL
INITIAL
 Unilaterally abort upon recovery

Failure in READY
 The coordinator has been informed
about the local decision
 Treat as timeout in READY state
and invoke the termination protocol

Failure in ABORT or COMMIT
Prepare
Vote-abort
Prepare
Vote-commit
READY
Global-abort
Ack
Global-commit
Ack
 Nothing special needs to be done
ABORT
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
COMMIT
Page 10-12. 114
2PC Recovery Protocols –
Additional Cases
Arise due to non-atomicity of log and message send
actions
 Coordinator site fails after writing “begin_commit”
log and before sending “prepare” command
 treat it as a failure in WAIT state; send “prepare”
command

Participant site fails after writing “ready” record in
log but before “vote-commit” is sent
 treat it as failure in READY state
 alternatively, can send “vote-commit” upon recovery

Participant site fails after writing “abort” record in
log but before “vote-abort” is sent
 no need to do anything upon recovery
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 115
2PC Recovery Protocols –
Additional Case

Coordinator site fails after logging its final
decision record but before sending its decision to
the participants
 coordinator treats it as a failure in COMMIT or
ABORT state
 participants treat it as timeout in the READY state

Participant site fails after writing “abort” or
“commit” record in log but before
acknowledgement is sent
 participant treats it as failure in COMMIT or ABORT
state
 coordinator will handle it by timeout in COMMIT or
ABORT state
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 116
Problem With 2PC

Blocking
 Ready implies that the participant waits for the
coordinator
 If coordinator fails, site is blocked until recovery
 Blocking reduces availability


Independent recovery is not possible
However, it is known that:
 Independent recovery protocols exist only for single
site failures; no independent recovery protocol exists
which is resilient to multiple-site failures.

So we search for these protocols – 3PC
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 117
Three-Phase Commit


3PC is non-blocking.
A commit protocols is non-blocking iff
 it is synchronous within one state
transition, and
 its state transition diagram contains
no state which is “adjacent” to both a commit
and an abort state, and
 no non-committable state which is “adjacent”
to a commit state



Adjacent: possible to go from one stat to
another with a single state transition
Committable: all sites have voted to
commit a transaction
 e.g.: COMMIT state
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 118
State Transitions in 3PC
Coordinator
INITIAL
INITIAL
Commit command
Prepare
Prepare
Vote-commit
Prepare
Vote-abort
WAIT
Vote-abort
Global-abort
ABORT
READY
Vote-commit
Prepare-to-commit
PRECOMMIT
Ready-to-commit
Global commit
Global-abort
Ack
ABORT
Prepared-to-commit
Ready-to-commit
PRECOMMIT
Global commit
Ack
COMMIT
Distributed DBMS
Participants
© 1998 M. Tamer Özsu & Patrick Valduriez
COMMIT
Page 10-12. 119
Communication Structure
P
P
P
P
P
P
C
C
C
P
P
P
P
P
P
ready?
yes/no
Phase 1
Distributed DBMS
C
pre-commit/
pre-abort?
yes/no
Phase 2
© 1998 M. Tamer Özsu & Patrick Valduriez
commit/abort
ack
Phase 3
Page 10-12. 120
Site Failures –
3PC Termination
Coordinator
INITIAL

 Who cares
Commit command
Prepare

ABORT

Timeout in PRECOMMIT
Vote-commit
Prepare-to-commit
 Participants may not be in
PRECOMMIT
Ready-to-commit
Global commit
COMMIT
Distributed DBMS
Timeout in WAIT
 Unilaterally abort
WAIT
Vote-abort
Global-abort
Timeout in INITIAL
PRE-COMMIT, but at least in
READY
 Move all the participants to
PRECOMMIT state
 Terminate by globally
committing
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 121
Site Failures –
3PC Termination
Coordinator
INITIAL
Commit command
Prepare

 Just ignore and treat the
WAIT
Vote-abort
Global-abort
ABORT
Timeout in ABORT or
COMMIT
Vote-commit
Prepare-to-commit
PRECOMMIT
transaction as completed
 participants are either in
PRECOMMIT or READY
state and can follow their
termination protocols
Ready-to-commit
Global commit
COMMIT
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 122
Site Failures –
3PC Termination
Participants

INITIAL
Timeout in INITIAL
 Coordinator must have
failed in INITIAL state
 Unilaterally abort
Prepare
Vote-commit
Prepare
Vote-abort

READY
Global-abort
Ack
ABORT
 Voted to commit, but does
not know the coordinator's
decision
 Elect a new coordinator
and terminate using a
special protocol
Prepared-to-commit
Ready-to-commit
PRECOMMIT

Global commit
Ack
COMMIT
Distributed DBMS
Timeout in READY
Timeout in PRECOMMIT
 Handle it the same as
timeout in READY state
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 123
Termination Protocol Upon
Coordinator Election
New coordinator can be in one of four states: WAIT,
PRECOMMIT, COMMIT, ABORT
 Coordinator sends its state to all of the participants asking
them to assume its state.
 Participants “back-up” and reply with appriate messages,
except those in ABORT and COMMIT states. Those in these
states respond with “Ack” but stay in their states.
 Coordinator guides the participants towards termination:
If the new coordinator is in the WAIT state, participants can be in
INITIAL, READY, ABORT or PRECOMMIT states. New
coordinator globally aborts the transaction.
 If the new coordinator is in the PRECOMMIT state, the
participants can be in READY, PRECOMMIT or COMMIT states.
The new coordinator will globally commit the transaction.
 If the new coordinator is in the ABORT or COMMIT states, at the
end of the first phase, the participants will have moved to that
state as well.

Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 124
Site Failures – 3PC Recovery

Coordinator
INITIAL
Failure in INITIAL
 start commit process upon
recovery
Commit command
Prepare

Failure in WAIT
 the participants may have
WAIT
Vote-abort
Global-abort
ABORT
Vote-commit
Prepare-to-commit
PRECOMMIT
Ready-to-commit
Global commit

COMMIT
Distributed DBMS
elected a new coordinator and
terminated the transaction
 the new coordinator could be
in WAIT or ABORT states 
transaction aborted
 ask around for the fate of the
transaction
Failure in PRECOMMIT
 ask around for the fate of the
transaction
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 125
Site Failures – 3PC Recovery
Coordinator
INITIAL
Commit command
Prepare

 Nothing special if all the
WAIT
Vote-abort
Global-abort
ABORT
Failure in COMMIT or
ABORT
Vote-commit
Prepare-to-commit
acknowledgements have been
received; otherwise the
termination protocol is
involved
PRECOMMIT
Ready-to-commit
Global commit
COMMIT
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 126
Site Failures – 3PC Recovery

Participants
INITIAL
Failure in INITIAL
 unilaterally abort upon
recovery
Prepare
Vote-commit
Prepare
Vote-abort

Failure in READY
 the coordinator has been
READY
Global-abort
Ack
Prepared-to-commit
Ready-to-commit 
informed about the local
decision
 upon recovery, ask around
Failure in PRECOMMIT
 ask around to determine how
ABORT

Global commit
Ack
COMMIT
Distributed DBMS
the other participants have
terminated the transaction
PRECOMMIT
Failure in COMMIT or
ABORT
 no need to do anything
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 127
Network Partitioning

Simple partitioning
 Only two partitions

Multiple partitioning
 More than two partitions

Formal bounds (due to Skeen):
 There exists no non-blocking protocol that is
resilient to a network partition if messages are
lost when partition occurs.
 There exist non-blocking protocols which are
resilient to a single network partition if all
undeliverable messages are returned to sender.
 There exists no non-blocking protocol which is
resilient to a multiple partition.
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 128
Independent Recovery Protocols
for Network Partitioning

No general solution possible
 allow one group to terminate while the other is
blocked
 improve availability

How to determine which group to proceed?
 The group with a majority

How does a group know if it has majority?
 centralized

whichever partitions contains the central site should
terminate the transaction
 voting-based (quorum)

Distributed DBMS
different for replicated vs non-replicated databases
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 129
Quorum Protocols for
Non-Replicated Databases

The network partitioning problem is
handled by the commit protocol.

Every site is assigned a vote Vi.

Total number of votes in the system V

Abort quorum Va, commit quorum Vc
 Va + Vc > V where 0 ≤ Va , Vc ≤ V
 Before a transaction commits, it must obtain
a commit quorum Vc
 Before a transaction aborts, it must obtain an
abort quorum Va
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 130
State Transitions in
Quorum Protocols
Coordinator
INITIAL
INITIAL
Commit command
Prepare
Prepare
Vote-abort
WAIT
Vote-abort
Prepare-to-abort
PREABORT
Ready-to-abort
Global-abort
Distributed DBMS
Prepare
Vote-commit
READY
Vote-commit
Prepare-to-commit
PRECOMMIT
Ready-to-commit
Global commit
ABORT
Participants
COMMIT
Prepared-to-abortt
Ready-to-abort
PREABORT
Prepare-to-commit
Ready-to-commit
PRECOMMIT
Global-abort
Ack
ABORT
© 1998 M. Tamer Özsu & Patrick Valduriez
Global commit
Ack
COMMIT
Page 10-12. 131
Quorum Protocols for
Replicated Databases


Network partitioning is handled by the
replica control protocol.
One implementation:
 Assign a vote to each copy of a replicated data
item (say Vi) such that i Vi = V
 Each operation has to obtain a read quorum (Vr)
to read and a write quorum (Vw) to write a data
item
 Then the following rules have to be obeyed in
determining the quorums:
Distributed DBMS

Vr + Vw > V
a data item is not read and written
by two transactions concurrently

Vw > V/2
two write operations from two
transactions cannot occur
concurrently on the same data item
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 132
Use for Network Partitioning

Simple modification of the ROWA rule:
 When the replica control protocol attempts to read
or write a data item, it first checks if a majority of
the sites are in the same partition as the site that
the protocol is running on (by checking its votes).
If so, execute the ROWA rule within that
partition.

Assumes that failures are “clean” which
means:
 failures that change the network's topology are
detected by all sites instantaneously
 each site has a view of the network consisting of
all the sites it can communicate with
Distributed DBMS
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 133
Open Problems

Replication protocols
 experimental validation
 replication of computation and communication

Transaction models
 changing requirements
cooperative sharing vs. competitive sharing
 interactive transactions
 longer duration
 complex operations on complex data

 relaxed semantics

Distributed DBMS
non-serializable correctness criteria
© 1998 M. Tamer Özsu & Patrick Valduriez
Page 10-12. 134
Transaction Model Design Space
Object complexity
active
objects
ADT +
complex
objects
ADT
instances
simple
data
flat
Distributed DBMS
closed
nesting
open
mixed
nesting
© 1998 M. Tamer Özsu & Patrick Valduriez
Transaction
structure
Page 10-12. 135